Frame-relay is one of the WAN protocols you need to understand if you plan to become CCNA

certified. Its also one of the most difficult protocols to understand for most CCNA students. In this
article Ill explain to you why we use(d) frame-relay and how it works. Before we start looking at
frame relay let me tell you a little story. We have a network with four sites:

Theres New York, Los Angeles, Miami and Boston. Since we want connectivity between the sites we
have an ISP who sold us three leased lines:

Between New York and Los Angeles.

Between New York and Miami.

Between New York and Boston.

Using leased lines is awesome! You are the only person using these lines since you are paying for
them. This means high quality and probably a low chance of congestion (if you have fast links). Its
also pretty secure since its only your traffic thats flowing through these lines.
There are also a few downsides to using leased lines however:

You are the only one using these lines so you gotta pay for them, being exclusive is fun but
expensive.

On router New York youll need three interfaces for each leased line, more interfaces means
more money.

What happens if you are going to move the site in New York? Its not always possible to
move the leased lines with you.
What if the New York router crashes? Boston, Miami and Los Angeles will be isolated as

well.

This is a picture of frame relay and it works a bit different. The idea behind frame relay is that you
have a single infrastructure from the service provider and multiple customers are connected to it,
effectively sharing everything.
In the middle you see a cloud with an icon you probably havent seen before. This icon is the frame
relay switch. The cloud is called the frame relay cloud and the reason it has this name is because
for us as customer its unknown what happens in the frame relay cloud. This is the service providers
infrastructure and we really dont care what happens therewe are the customer and all we want is
connectivity!
What else do you see? There are two customers (A and B) and each of them has a HQ
(Headquarters) and a branch office.
One more picture, heres a frame relay network with three routers from one company:

Theres a router at the headquarters and we have two branch offices. All of them are connected to
the frame relay cloud.

We call our service provider since we want connectivity and the first question theyll ask us is which
sites should be connected?. In the example above you can see two virtual circuits, the red and blue
one. With frame relay theres a difference between the physical and logical connections. The physical
connection is just the serial cable which is connected to the provider. Our logical links are virtual
circuits. As you can see there is a virtual circuit from branch Office 1 to the HQ router and another
one from branch office 2 to the HQ routers.
This means that we can send traffic through our virtual circuits between:

Branch office 1 and HQ.

Branch office 2 and HQ.

There is no virtual circuit between branch office 1 and branch office 2. Does this mean there is no
connectivity between them? No you can still have connectivity between them by sending data to the
HQ router! Of course you can get another virtual circuit between branch office 1 and branch office 2
but youll have to pay for it. Virtual circuits are also called PVC (Permanent Virtual Circuit).
You also pay for a certain speed called the CIR (Committed Information Rate). The cool thing
about frame relay is that when no other customers are using the frame relay network its possible you
get a higher speed than what you paid forthe CIR however is a speed that is guaranteed.
How do we know if a PVC is working or not?

Frame-relay uses something called LMI which stands for Local Management Interface. LMI has
two functions:

Its a keepalive mechanism.

It tells us if the PVC is active or inactive.

It also gives us a DLCI (Data Link Connection Identifier). Ill get back to this in a bit
There are 3 types of LMI. They all do the same thing but there are three standards which are not
compatible with each other. Whatever you choose make sure its the same between two devices:

Cisco

ANSI T1.617 Annex D

ITU-T Q.933 Annex A

So if you pick Cisco on one side, use Cisco on the other side as well.

Heres an example of LMI in action. In the middle we have the frame relay switch. LMI packets are
sent between Router A and the frame relay switch and router B and the frame relay switch. The
frame relay switch tells our routers that the PVC is active.
What else do you need to know about frame relay? Let me show you the OSI model:

WAN protocols describe the physical (layer 1) and data link (layer 2). What does frame relay use on
the data link layer? We dont use MAC addresses since thats Ethernet but we do have something
else called a DLCI (Data Link Connection Identifier).

For each PVC you will get a DLCI per router. In our example above you can see that for the PVC
between router HQ and branch office 1 we have DLCI 102 on the HQ router and DLCI 201 on the
branch office 1 router.
Between router HQ and router branch office 2 we have DLCI 103 on HQ and DLCI 301 on branch
office 2. Our DLCI is nothing more but a unique identifier for the data link layer per PVC.

Now there is an important concept to grasp and remember about DLCI. DLCIs are only locally
known to the router! Your router does not know the DLCI of the router on the other side. This is
different if you compare it to Ethernet. In our Ethernet world you need to know the MAC address of
the computer on the other side in order to send something to it.
This is just like taking a train. If you are at the train station you walk to the correct train platform and
take the train. You have no idea on which train platform you will arrive and you dont care.
Frame-relay supports multiple topologies:

Full-mesh

Partial-mesh

Hub and Spoke

Let me show you an example of each topology type:

This is our full-mesh topology. As you can see there is a PVC between every router.

This is partial-mesh. The more important routers will have multiple connections to others.

This is the hub and spoke model. The router on the left is our hub and the other routers are spokes.
If the spokes want to communicate with each other theyll have to send traffic towards the hub router.
Frame-relay is NBMA (non-broadcast multi-access)
Keep this in mind. What it means is that frame relay is multi-access since all routers can access the
network but you are unable to send broadcasts over the frame relay network. No broadcast also
means you are unable to send multicast traffic. No multicast means youll be in trouble with routing
protocols. Rip version 2, OSPF and EIGRP all use multicast. Does this mean you cant use routing
protocols with frame relay? Well no but its a bit tricky:

RIP, OSPF and EIGRP can also use unicast instead of multicast.

There is a method to emulate broadcasts over your frame relay network.

What other problems might we encounter with frame relay and routing? Do you remember the
characteristics of distance vector routing protocols (RIP and EIGRP)?
In the picture above I have configured RIP on all the routers. Router branch office 1 is sending
routing information towards router Customer HQ. If we look at the routing table we see this routing
information on router HQ.

Do you remember the split-horizon rule? Whatever you learn from your neighbor you dont advertise
back to them. To be more specific: whatever you learn on an interface you dont advertise it
back out on the same interface.
We are using two PVCs but on router HQ there is still only one physical interface. Split-horizon will
prevent the advertisement of routing information towards router branch office 2.
How can we solve this problem?

You can disable split horizon (the default on physical interfaces).

You can use sub-interfaces.

If you use a sub-interface you dont have the split-horizon problem since you are learning routing
information on serial0/0.1 and advertising it out of serial0/0.2
Frame-relay can use point-to-point sub-interfaces or point-to-multipoint sub-interfaces. If you use
point-to-point it will solve your split-horizon problem but youll need to use a different IP subnet per
PVC. Point-to-multipoint means you have the split-horizon problem but you can use a single IP
subnet for all PVCs.
Remember ARP (address resolution protocol)? When we use ARP for Ethernet we need to learn the
MAC address of the computer we want to send something to. ARP effectively maps the destination
IP address to the destination MAC address.
Frame-relay uses inverse ARP and is a bit different. Remember my story about the train platform
and how your router only knows its local DLCI? You dont know the DLCI of the other side. Inverse
ARP is going to map your local DLCI to the IP address of the other side:

Router Frodo in my example above has mapped the IP address of router Gandalf (192.16.12.2) to its
local DLCI 102. Thats inverse ARP.
Lets see it in more detail:

When we configure frame relay heres what happens:

1. Our router will do a status enquiry using LMI.
2. The frame relay switch will give us our DLCI number (or you can configure it yourself).
3. The routers send a hello message with their IP address.
4. The routers receive the hello message from the other side and create a map with the
remote IP address + local DLCI number.
Router Frodo will now know it can reach IP address 192.168.12.2 by sending traffic through the PVC
with DLCI 102. Router Gandalf will know that it can reach IP address 192.168.12.1 through the PVC
with DLCI 201.
========================================================
Download pdf version here http://tinyurl.com/pv7lqfr
========================================================

All the Best,

Maher Abdelshkour
Like